The Center for Security and Emerging Technology (CSET) offers the following submission for consideration by the Office of Science and Technology Policy; the NSTC Select Committee on Artificial Intelligence; the NSTC Machine Learning and AI Subcommittee; the National AI Initiative Office; and the NITRD National Coordination Office. OSTP, on behalf of its partners, requested input on updating the National Artificial Intelligence Research and Development Strategic Plan.
In this submission, CSET makes 15 recommendations to improve upon the National Artificial Intelligence Research and Development Strategic Plan (Strategic Plan):
- Adopt caution when pursuing long-term AI research that could generate “general purpose” or “human-like” artificial intelligence.
- Promote the development of tools for “trust calibration” to enable safe and effective human-AI collaboration.
- NSF and DARPA should fund privacy-preserving computer vision research as an alternative to automated mass facial surveillance.
- Promote research into attacks on AI systems that are more likely to resemble real-world threat scenarios.
- Dedicate resources to studying and mitigating software vulnerabilities in the AI supply chain.
- Research and develop AI specific standards and processes for assessing AI maturity.
- Pursue cybersecurity-relevant datasets and testbeds as a special area of focus.
- Increase the interoperability of existing data resources.
- Encourage the use of operationally relevant metrics and the evaluation of AI in the operational context and condition in which it will be used.
- Emphasize characterizing performance across use conditions.
- The United States should designate developing global facial recognition standards a new priority on its AI standards list, and incentivize U.S. companies’ participation in standards bodies.
- Define the AI R&D workforce, in addition to computer research scientists, and compile and publish data on the composition of this workforce.
- Identify and remedy inefficiencies in the immigration process that prevent foreign-born AI experts from obtaining residency in the United States.
- Locate gaps in private sector’s AI R&D agenda and forge public-private partnerships to target these areas.
- Make increasing transparency in the U.S. federal AI R&D ecosystem a new strategic priority.